Pytońska treść do mowy – Polish Text to Speech library for Python
Project description
PyTDM
ie. Pytońska treść do mowy which is Polish for Python Text to Speech. Package for turning text written in Polish into speech.
Both for standard Python3 and for iOS.
ok but why
This lil library was designed in order to assure that people programming for Pollacks have some sort of an offline-working text to speech python software.
For English speaking people there already is the pyttsx3
library which provides such functionalities.
If you want your programme to talk you simply run few commands
import pyttsx3
engine = pyttsx3.init()
engine.say("now we're talkin'")
engine.runAndWait()
and it's literally that easy! But for Polish there was no such thing. Until now.
installation
it is avalaible on PyPI! you can just run:
pip install pytdm
and you can use it just like that!
iOS
for the iOS version see the pythonista directory on github, do not use pip
there.
how it's made
It's based on the pyttsx3
package I've just mentioned. The thing is that contrary to what many English speaking people think (looking at u americans) Polish actually is pronounceable for someone knowing English.
What PyTDM
does is
- first it translates* the given polish string/text into english
- then it gets pronounced with the
pyttsx3
The asterisk above next to the word "translates" is there for a reason – it's not exactly translation. More like transcription or transliteration. We shall get back to that later
usage
For simple basic stuff you should do it as follows:
import pytdm
pytdm.mów("dzień dobry, dobranoc")
or the second way:
from pytdm import mowa
mowa.mów("dzień dobry, dobranoc")
and then you can happily listen to the sweet sound of the polish language spoken by the pyttsx3
synthesiser voice. Isn't that great?
saving files
With version 0.1.0
comes the functionality of saving spoken text in files! It is similar to simply making the computer say stuff:
pytdm.zapisz("dzień dobry świecie!", "example.mp3")
and your file will be saved in your cwd under the name "example.mp3"
little disclaimer
the functions have Polish names like mów
or tłumacz
with those funny strange letter but if you want you can use them without the diacritics eg. write mow
or tlumacz
. They will work just fine.
In general if have any problems first you can type eg. help(pytdm.mow)
and read the info provided there.
examples
An example for how well does the software work with approxima... translation of polish words is to be seen in the demo.py
file (avalaible on my github). You just can run it and then see how well it handles the most sacred polish song (actually the second sacred-est. For the most sacred one see barka.py
. Both demo files are in the demos
github directory) – the anthem of the Third Polish Republic.
There is a video showing how the demo.py
file works (recorded in low quality by me and posted on youtube) here.
Also with version 0.1.0
the french mode has been introduced therefore there's another demo file demo_fr.py
. A video of it being run is here
dependencies
all you need are built-in packages like re
and apart from that the pyttsx3
package (v>=2.7
). It is avalaible on pypi so you can just do the classic:
pip install pyttsx3
what OS?
The only problem is it's different for every OS.
- For all I know it works really well on macOS – I use this OS myself therefore I can easily test that there and adjust it respectively.
For other OS I have only some feedback from other people:
- Windows is stoopid (per usual) and reads 'ch' as /k/ and not /t͡ʃ/. Apart from that and maybe some other mistakes it is fine.
- it's not bad on Linux but it has problems with eg. consonant clusters like szcz. When it tries to pronounce them as shtch it spells (es aitch tee cee aitch...) the cluster instead of just just saying /ʃt͡ʃ/ as other versions tend to do. Also it sounds robotic which is w e i r d
In conclusion: it works ok for all of them but some sounds are realised differently on different OSs.
mobile versions – iOS
for iOS see the pythonista
directory on github. Everything there is the iOS version of PyTDM
.
Function names are weird?
As a cautious reader might have noticed the main speaking function is mów
(fyi it means say just like in the pyttsx3
). One could ask wthell? or more properly co do diabła? but that is exactly how the package was intended to be: the functions have polish names. Deal with it.
behind the scenes
Now for some calrification about how the so called translation process actually works.
For every word passed to the mów
function it is first tanslated by another polish-named func tłumacz
(ie. translate) and it calls 2 more functions first:
-
repolonizuj
ie. repolonise – it deals with all the weird polish ortographic stuff like the diagraphs, some consonants being devoiced etc. Futuryści inspired -
anglicyzuj
ie. anglicise – it takes the repolonised text and tries to find the closest approximations for all the sounds that are to be found. It's doing its best.
only then mów
gives the anglicised repolonised text to the engine.say
as shown above.
So when you do something like mów('czuję, że będzie dziś dość średni dzień')
it first calls tłumacz
which calls repolonizuj
that returns this simplified polish text:
'czuje, że bendźe dźiś dość średni dźen'
and it is passed to the anglicyzuj
which gives the final result to be said by pyttsx3
:
'choo yeah, zshehh behnjehh jeesh dawshtch shrehdnee jehn'
it sure is amazing.
French
As you may have known the pyttsx3
library offers any speech synthesiser avalaible to the system. Therefore it may happen one does also have the french one and now it is possible to use it with pytdm
too.
It is a relatively new feature so you shouldn't expect it to be flawless. As for now pronouncing numbers is not even implemented here. Also it hasn't been tested anywhere apart from macOS yet.
he good thing is it is actually easier to transcribe Polish into French than into English. Why? Well despite what many people think about the French language it is actually pretty logical and much more regular than English. Its reading rules are not that messy and there are not as many strange edge cases.
How to use the french mode?
You must have the pytdm
version 0.1.0
or higher installed.
just add the lang="fr"
argument or simply fr
when calling functions like mów
or tłumacz
. To set it back to english set lang="en"
.
>>> import pytdm
>>> pytdm.tłumacz("czuję, że będzie dziś dość średni dzień", "fr")
'tchouyé, jé baindjé djich dochtch chrédgni djègn'
>>> pytdm.mów("czuję, że będzie dziś dość średni dzień", "fr")
czuję, że będzie dziś dość średni dzień
And there is the function francyzuj
that works the same as anglicyzuj
mentioned before but for french
TO DO
- polish the Polish
- implement pronouncing numbers above 199
- better handling of numbers mixed-in with words
- no hardcoded words is the goal
- better handling of syllables, diphtongs and other strange edgecases eg. je, ej, aj, świe etc
- make french mode as good as the standard english one. right now there is no pronounciation for numbers in
- a good idea would be to provide alternative rules of translation for different operating systems in the future
- add numbers to the french variant
- some proper documentation for functions and the translation process. Regexp is famously complicated as hell so an entire file explaining each and everyone of all those patterns etc is not a bad idea.
- optional
gTTS
mode - extend this list
feel free to fork the github repo and provide some videos:)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.